How to share a single sequence between multiple SQL Server databases? - sql-server

Question

I need to share a single SQL Server sequence between tables in two different SQL Server databases. This example works just fine for two existing tables and a sequence defined on the same database:
USE DatabaseA ;
GO
CREATE SEQUENCE dbo.TransactionSequence START WITH 1 INCREMENT BY 1 ;
GO
ALTER TABLE Table1 ADD TransactionSequenceID int NOT NULL
CONSTRAINT DF_Table1_TransactionSequence DEFAULT (NEXT VALUE FOR dbo.TransactionSequence) ;
ALTER TABLE Table2 ADD TransactionSequenceID int NOT NULL
CONSTRAINT DF_Table2_TransactionSequence DEFAULT (NEXT VALUE FOR dbo.TransactionSequence) ;
GO
However, this command fails for a table in a different database:
USE DatabaseB ;
GO
ALTER TABLE Table3 ADD TransactionSequenceID int NOT NULL
CONSTRAINT DF_Table3_TransactionSequence DEFAULT (NEXT VALUE FOR DatabaseA.dbo.TransactionSequence) ;
GO
It fails because SQL Server does not allow a DEFAULT constraint to reference a sequence in another database.
A sequence cannot be referenced indirectly through a synonym in this context, and NEXT VALUE FOR <sequence> is not allowed inside of a function or several other possibly-useful constructs. Is it possible for tables in different databases to share a single sequence? If so, then how?

Answer 1

I've implemented something like this in the past and depending on your need/use case it might work for you. Let's say that you have need for a sequence that generates WidgetID values in multiple databases. You'd do it something like this:
use [Widgets_001];
create sequence SQ_WidgetID
as int
start with 1
minvalue 1
maxvalue 1000000
increment by 1;
use [Widgets_002];
create sequence SQ_WidgetID
as int
start with 1000001
minvalue 1000001
maxvalue 2000000
increment by 1;
Once you have the sequences created, you're allowed to use them in whatever default constraints you want.
The idea here being that they allocate from disjoint ranges. Because the sequences allocate independently, the databases can even be on different servers. This does come with the cost of needing to monitor for range exhaustion (e.g. when the sequence in Widget_001 is over 950,000 it's probably time to start thinking about allocating a new range) and care in re-seeding the sequence to allocate from a new range (which isn't too bad if you have a way to look at all of the sequences across databases and always allocate new ranges to have minvalue strictly larger than the max of all of the existing maxvalues).
I'd personally resist the temptation to attribute any meaning to the values generated by these respective sequences. That is, it's easy to say "Well... that's WidgetID 1234. Therefore it's in Widgets_001.". That comes with its own problems.